Multivariate methods for monitoring structural change

Detection of structural change is a critical empirical activity, but continuous 'monitoring' of time series for structural changes in real time raises well-known econometric issues. These have been explored in a univariate context. If multiple series co-break, as may be plausible, then it is possible that simultaneous examination of a multivariate set of data would help identify changes with higher probability or more rapidly than when series are examined on a case-by-case basis. Some asymptotic theory is developed for a maximum CUSUM detection test. Monte Carlo experiments suggest that there is an improvement in detection relative to a univariate detector over a wide range of experimental parameters, given a sufficiently large number of co-breaking series. The method is applied to UK RPI inflation in the period after 2001. A break is detected which would not have been picked up by univariate methods.